Re: polymix-4 workload time duration

From: Alex Rousskov (rousskov@measurement-factory.com)
Date: Thu Dec 04 2003 - 22:19:36 MST

Next message: Vittorio Lindia: "reduce IP addresses"
Previous message: Alex Rousskov: "Re: Understanding Histograms"
In reply to: Henrik Nordstrom: "Re: polymix-4 workload time duration"
Next in thread: Charlie Younghusband: "Understanding Histograms"
Messages sorted by: [ date ] [ thread ] [ subject ] [ author ]

Three additional bits of information regarding shortening tests
to get PolyMix-4 estimates:

        - Some vendor labs have found product-specific workloads
          and fudge factors to get good PolyMix-4 estimates from
          shorter tests (with full dirty cache disks). If you
          test a single family of products, you may be able to
          find such shortcuts as well.

        - With Polygraph 2.8.0, you can reuse working set
          from a previous tests. With some effort, this can
          shorten consecutive similar PolyMix-4 runs to
          just a few hours each.
          http://www.web-polygraph.org/docs/userman/recycle.html

        - With Polygraph 2.8.0, you can have a watchdog-controlled
          test that attempts to find PolyMix-4 sustained peak rate
          by varying load based on runtime measurements
          such as response time or the number of errors
          (we often call it a Peak Finder approach):
          http://www.web-polygraph.org/docs/userman/inflight.html

All these timesaving tricks require clean, full PolyMix-4 runs
as a final confirmation of estimates, of course.

HTH,

Alex.

-- 
Protocol performance, functionality, and reliability testing.
Tools, services, and know-how.
http://www.measurement-factory.com/
On Tue, 2 Dec 2003, Henrik Nordstrom wrote:
> On Sun, 5 Oct 2003, Muhammad Talha wrote:
>
> > I want to confirm that minimum polymix-4 workload test time duration is appox 11 hours ?
>
> Sounds reasonable. But the length of the fill phase depends on your cache
> size and fill request rate so it varies slightly depending on your
> configuration.
>
> The test phase alone is 20m+4h+20m+20m+20m+4h+20m or 9 hours. But before
> the test phase begins it fills the cache with 2 * cache_size of cacheable
> content with a 5% hit ratio and a rate of FillRate.
>
> > should I run the test for few  hours I want or they will useless results
>
> There is no meaningful results until the top1 phase, and this needs to
> have been running for a significant amount of time before the results is
> somewhat useable.
>
> > should i have to stop the test manually or it will automatically terminated
> > after completion
>
> The test terminates automatically when completed.
>
> Regards
> Henrik
>
>

Next message: Vittorio Lindia: "reduce IP addresses"
Previous message: Alex Rousskov: "Re: Understanding Histograms"
In reply to: Henrik Nordstrom: "Re: polymix-4 workload time duration"
Next in thread: Charlie Younghusband: "Understanding Histograms"
Messages sorted by: [ date ] [ thread ] [ subject ] [ author ]

This archive was generated by hypermail 2b29 : Mon Feb 06 2006 - 12:00:27 MST